Robust Separation-Enhanced NRC Method for Multiple Periodicity Detection: Applications in Bearing Compound Fault Diagnosis

Identification of compound faults in rotating machinery bearings is crucial for ensuring reliability and safety. Traditional methods face challenges in detecting weak multi-periodic components amidst strong noise during bearing malfunctions. While the noise-resistant correlation (NRC) method excels in single hidden period detection, it struggles with unclear peaks under strong noise and complex fault diagnoses. This paper introduces a novel approach, the Separation-Enhanced NRC method, which addresses these challenges. Firstly, we approach the NRC method from a different perspective and propose an enhanced version, referred to as the enhanced NRC (E-NRC) method, which amplifies the magnitude of peaks at periodic locations. Secondly, we integrate maximum overlap discrete wavelet packet transform (MODWPT) and construct the Anti-Gini index as a metric to evaluate the magnitude of periodic components in the decomposed signal, facilitating multi-period detection. In addition, a thresholding method is proposed to filter components containing periodic information. The effectiveness of the proposed method in detecting compound faults has also been tested by simulation and experiment studies.